Journal of Environmental Management 207 (2018) 292e302

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

Research article Comparison of trailside degradation across a gradient of trail use in the Sonoran Desert

* Helen Ivy Rowe a, , Melanie Tluczek a, Jennifer Broatch b, Dan Gruber a, Steve Jones a, Debbie Langenfeld a, Peggy McNamara a, Leona Weinstein a a McDowell Sonoran Conservancy, Scottsdale, AZ, USA b School of Mathematical and Natural Sciences, State University at the West Campus, Glendale, AZ, USA article info abstract

Article history: As recreational visitation to the Sonoran Desert increases, the concern of scientists, managers and ad- Received 16 June 2017 vocates who manage its natural resources deepens. Although many studies have been conducted on Received in revised form trampling of undisturbed vegetation and the effects of trails on adjacent and communities, 13 October 2017 little such research has been conducted in the arid southwest. We sampled nine 450-m trail segments Accepted 10 November 2017 with different visitation levels in Scottsdale's McDowell Sonoran Preserve over three years to understand the effects of visitation on soil , trailside soil crusts and plant communities. Soil crust was reduced by 27e34% near medium and high use trails (an estimated peak rate of 13e70 visitors per hour) Keywords: Trail impacts compared with control plots, but there was less than 1% reduction near low use trails (peak rate of two to Native plant species four visitors per hour). We did not detect soil erosion in the center 80% of the trampled area of any of the Nonnatives trails. The number of perennial plant species dropped by less than one plant species on average, but Soil crust perennial plant cover decreased by 7.5% in trailside plots compared with control plots 6 m off-trail. At the Recreation current levels of visitation, the primary management focus should be keeping people on the originally Sonoran Desert constructed trail tread surface to reduce impact to adjacent soil crusts. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction campsites, or b) testing resistance and resilience of undisturbed areas through controlled trampling in undisturbed areas (Monz An estimated 78.3 million adults visited trails in the United et al., 2013). States in 2008 and that number is predicted to increase 30% by On established trails, occurs initially at the time of 2030 (White et al., 2014). One early comprehensive review esti- trail construction as a result of opening canopies by vegetation mated that trails and campsites together disturb only 1% of the total removal, compaction of soil and alteration of drainage patterns by area of wilderness (Cole, 1987), but this was based on estimates of removal of upper soil horizons, and modification of micro topog- area rather than direct experiments on the effects of disturbance on raphy, affecting microclimate (Cole, 1987). Subsequently, ongoing ecological function (Adkinson and Jackson, 1996). The concern over trail visitation has direct (trampling) and indirect effects (com- visitor disturbance in natural areas has motivated extensive pacted , reduced organic matter and reduced soil nutrient research on trail impacts, as seen by multiple review papers over changes) on trailside vegetation (Monz et al., 2010). As demand for the years (Ballantyne and Pickering, 2015; Cole, 1987; Hammitt and recreational trails and trail use continues to grow, understanding Cole, 1998; Kuss et al., 1990; Leung and Marion, 2000; Monz et al., anthropogenic environmental impacts will become increasingly 2013), and more recently, in research studying the effects of trails important to inform sustainable resource management. on surrounding ecological communities (as reviewed by Monz The Sonoran Desert is a prime destination for outdoor recrea- et al., 2013). Trail impact studies generally focus either on: a) tion. For example, Scottsdale's 13,000 ha McDowell Sonoran Pre- examining the changes along established trails and around serve (MSP), which is the largest urban preserve in the , received 750,000 visits in 2016, yet it is only one of many natural areas surrounding the Phoenix metropolitan area. Since environmental factors such as and geology, and the inter- * Corresponding author. mediate elements of topography, soil, and vegetation type E-mail address: [email protected] (H.I. Rowe). https://doi.org/10.1016/j.jenvman.2017.11.028 0301-4797/© 2017 Elsevier Ltd. All rights reserved. H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302 293 significantly affect the degree and type of trail degradation (Leung nearest weather stations during the study years 2014e2016 (Flood and Marion, 1996), each ecological system responds differently to Control District of Maricopa County, 2017). means trail visitation and associated impacts. In a recent review of 59 from the four nearest precipitation gauges indicated rainfall was original research papers on trail impacts, over 50% of the papers above average in the years preceding the first two sampling periods focused on just three types: temperate forests, alpine and of the study (27.7 cm in 2013, 26.9 cm in 2014; Table 1), and slightly montane grasslands and shrublands, and Mediterranean forests, below average rainfall in the year preceding the final sampling woodlands and sclerophyll scrub, while only 2 papers (3%) focused season (22.6 cm in 2015; Table 1). The Sonoran Desert climate in- on deserts and xeric shrublands (Ballantyne and Pickering, 2015), cludes two distinct rainy seasons: one in the winter (Decem- and these latter papers were either focused on the impacts of roads bereMarch), and one in the summer (JuneeSeptember). (Brooks and Lair, 2005), or on the effects of vehicles or pedestrian The MSP is topographically and biologically diverse, ranging in use on untrampled dune systems (Rickard et al., 1994). elevation from 515 to 1237 m above sea level. A biological inventory Soil crust is a key biological indicator of disturbance in south- conducted between 2011 and 2013 found 368 plant species and 188 west arid environments (Allen, 2009; Belnap, 1998). In the south- vertebrate animal species (Jones and Hull, 2014; McDowell Sonoran west and similar environments, biological soil crust plays an Conservancy, 2014). There are 14 distinct plant associations important role in increasing soil stability, water infiltration, and soil distributed across the MSP (Jones and Hull, 2014). Bedrock geology fertility in otherwise erodible, dry, and infertile soils (Belnap, 1994; and soil types differ across the MSP, with predominantly meta- Belnap and Gardner, 1993; Harper and Marble, 1988; Johansen, morphic rock in the south and decomposed granite in the north 1993; Metting, 1991; Williams et al., 1995). Consequently, soil (Skotnicki, 2016). crust loss can result in soil erosion and loss of soil nutrients (Belnap Three blocks were identified which contained similar attributes and Gillette, 1997; Harper and Marble, 1988; Schimel et al., 1985). within each block but were distinct between blocks (Table 2). Soil crusts are likely susceptible to trail impacts because they are Within each block, 3 trail segments were selected to represent a brittle when dry and crush easily with trampling (Belnap and gradient of trail visitation levels and to minimize differences in Gardner, 1993). plant association, soils, geology, slope, and elevation within block In our informal review of over 75 related papers, none of the (Table 2). A fourth control transect that had no visitation was trail impact studies in arid regions studied effects across a gradient established within each block at least 100 m from the trails. of use levels. Studies in other regions which did measure impact as Plant communities differed by block (Table 2). The common a function of use levels produced a variety of results (Ballantyne associated perennial plant species at the Gateway block are brit- and Pickering, 2015). Most trampling studies reported a definite, tlebush (Encelia farinosa), barrel cactus (Ferocactus cylindraceus), often curvilinear or asymptotic, positive relationship between buckhorn cholla (Cylindropuntia acanthocarpa), catclaw acacia increased use intensity and increased physical and biological im- (Acacia greggii), chain fruit cholla (Cylindropuntia fulgida), creosote pacts (Andres-Abell an et al., 2006; Ballantyne and Pickering, 2015; bush (), and saguaro cactus (Carnegiea gigantea). Boucher et al., 1991; Wimpey and Marion, 2010), but others found The Tom's Thumb block shares many of the common perennial no clear relationship with use levels (Nepal and Way, 2007) or that plant species with the Gateway block, but being higher elevation plant cover increased closer to trails, regardless of use intensity also contains Arizona desert ethorn (Lycium exsertum), Wright's (Bright, 1986; Hall and Kuss, 1989). Some studies found that other buckwheat (Eriogonum wrightii), fairy duster (Calliandra eriophylla), factorsdvisitor behavior, type of use (equestrian, bicycle, pedes- globe mallow (Sphaeralcea ambigua), goldeneye ( parishii), trian), floristic community, topography, and othersd appeared to and Mormon tea (Ephedra aspera). Creosote bush (Larrea tri- be more important than use levels in causing physical and biolog- dentata), and saguaro cactus (Carnegiea gigantea) are present in ical impacts (Adkinson and Jackson, 1996; D'Antonio et al., 2016; lower densities than they are in the Gateway block. Trail segments Dixon et al., 2004). within the Brown's Ranch block share the same common perennial Recreation activities can cause impacts to soil, vegetation, as Tom's Thumb block, however, saguaros are present in wildlife, and water (Leung and Marion, 1996), yet principles of greater density (Jones, 2015). sustainable natural resource management include preserving bio- logical diversity and providing safe, enjoyable experiences for vis- 2.2. Trail visitation levels itors. In order to balance these objectives in a given ecological system, it is critical to understand how increased use affects The research objectives were met by placing paired plots adja- biodiversity. We investigated the resilience of trailside vegetation cent and 6 m away from trails of contrasting visitation levels. Trail and soil crusts to different visitation levels in the Sonoran Desert. segments were chosen to capture low, medium and high visitation Thus, we had two objectives for our study: 1) to evaluate trail im- levels within the same biotic community and soil type according to pacts on vegetation and soil crusts, and 2) investigate whether estimates informed by preliminary data from mechanical and these impacts are affected by different levels of trail visitation in the volunteer counters. Sonoran Desert. We used two approaches to quantify visitation rates during the study. First, mechanical counters (Diamond Traffic Products, Model 2. Methods TTC-4420) were placed at the beginning of each trail segment transect (Fig. 1) and collected data for 24 months (2014e2015). 2.1. Sites Secondly, volunteers seated next to each mechanical counter counted trail visitors for 2 h during peak visitation (7e9 a.m. in The MSP is comprised of Sonoran Desert Upland habitat (Brown summer, 8e10 a.m. in spring and fall, and 9e11 a.m. in winter) on et al., 1979) which lies at the northeastern edge of the Phoenix the third Saturday of each month for January through November metropolitan area in central Arizona (33.59 N, 111.76 W). Due to the 2014 and 2015 (hereafter designated as “volunteer count”). proximity to the Phoenix urban core, the MSP receives heavy visi- The mechanical counters provide hourly counts of visitors tation by hikers, bikers, and equestrians. The City of Scottsdale throughout the year. Unfortunately, the mechanical counters estimates that there were approximately 750,000 individual visits appeared to have numerous data inconsistencies, including 1) in 2016. Motorized vehicles are not permitted in the MSP. daytime and nighttime hours sometimes were reversed, as evi- Annual temperatures ranged from 20 C to 46.7 C at the two denced by visitation occurring during the night (when the MSP is 294 H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302

Table 1 Mean annual precipitation (mm) data from the Flood Control District of Maricopa County (2017) weather stations closest to study sites.

Weather station # Weather stations Total 2013 Dev.a Total 2014 Dev. Total 2015 Dev. Long term meanb

4585 Reata pass wash 274.07 66.04 290.32 82.30 210.57 2.54 208.03 4595 Pinnacle pk vista 228.60 26.92 266.19 64.52 228.35 26.67 201.68 4935 Reata pass dam 335.28 80.77 287.02 32.51 248.41 6.10 254.51 5930 Fraesfield Mtn 316.23 15.24 278.89 22.10 267.97 33.02 300.99 Means 288.54 47.24 280.61 39.31 238.82 2.48 241.30

a Dev indicates deviation from the long term average. b Long term averages for rainfall guages began in different years: 4585 in 2001; 4595 in 1998; 4935 in 1993; 5930 in 1989. closed) on popular trails instead of during the day; 2) data were same side of the trail, a paired 1 m 1 m control plot was placed missing from some counters due to mechanical failure; 3) days of 6 m from the trail (5 m from the trailside plot), perpendicular to the the week were not accurate as evidenced by charts of visitation trail. showing peak visitation during the week (rather than on week- We measured trail depth to evaluate trail soil erosion at the ends) and finally, 4) erratic data sometimes were recorded with beginning and end of each transect between the two marker posts. unusually high numbers inconsistent with the trend for that trail. We created notches on each marker post at about 1 m height. To Rather than attempting to correct the data, we reported the establish the three sampling points, we tied a static 2 mm twisted mechanical counter results for the same 2 h blocks that were nylon string (>100# tensile strength) between the notches at each counted by volunteers one Saturday per month, using only months post so that the string was taut and level across the transect. The for which the mechanical counter data were complete during the trail center point and points near each edge of the trail were same times as the volunteer counts. In doing so, we were able to selected, so that all three points fell within the center 80% of the avoid many of the data errors described above. The mechanical treaded area of the trail. The distances of these three points from counts and the volunteer counts were divided by two to show an the upslope marker post were recorded and the same points were hourly average. re-sampled each year. Sampling was conducted by measuring The mechanical block counts were consistent with the volunteer vertically downward from the three sample points to the trail tread block counts. We used the volunteer block data as the basis for surface and recording the distances. visitation rate of the trails because we were most confident of the In the second year, we added new site control plots 100 m from a quality of this data, and because it had a more complete data set trail (no visitation) to test if the 6 m was a sufficient distance to than the mechanical block. control for trail disturbance effects on the trailside plots. If 100 m plots were similar to 6 m plots compared to the trailside plots, this would indicate that the 6 m distance was sufficient. One site control 2.3. Sampling design transect without paired plots was randomly placed in each block at least 100 m from the nearest trail. As in trail segments, 1 m 1m We designed the experiment to compare vegetation adjacent to quadrats were placed every 30 m for a total of 15 plots per site the trail (15 “trailside plots”) with control plots 6 m from the trail control (3 blocks 15 plots ¼ 45 plots total). Permanent markers (15 “6 m plots”) (3 trail segments x 30 plots 3 blocks ¼ 270 plots, were placed in all four corners of each quadrat to ensure the same Fig. 1) on trail segments across a gradient of visitation levels. By area would be sampled each year. using this design, the differences between trail and off-trail plots can be attributed to the construction and use of the trail (Leung and Marion, 1996). Trail segments were selected based on three criteria: 2.4. Sampling capturing 450 m between trail junctions, in the same floristic community, with generally consistent cross-trail slope for drainage. Visual estimates of coverage for each plant species and for Two marker posts were placed approximately 3 m from trail center visible soil crust were made in each 1 m 1 m plot using six cover on either side of the trail at the beginning and end of the 450 m classes modified from the Braun-Blanquet cover-abundance scale transect. For trailside plots, 1 m 1 m quadrats were placed every (Mueller-Dombois and Ellenberg, 1974) as follows (cover 30 m so that one side of the quadrat frame was lined up with the class ¼ range ¼ midpoint): 1¼ <1% ¼ 0.5; 2 ¼ 1e5% ¼ 3; most visually obvious perimeter of the trail area which receives the 3 ¼ 5e25% ¼ 15; 4 ¼ 25e50% ¼ 37.5; 5 ¼ 50e75% ¼ 62.5; majority of the traffic (95%). The boundary is defined by areas of 6 ¼ 75e100% ¼ 87.5. Plants were considered in the plot if any part apparent trampling, disturbance to organic litter, and where of the plant was in or hanging over the edge of the quadrat. Plant vegetation cover is reduced or absent (Marion and Leung, 2001; and soil crust sampling was conducted at peak spring annual Olive and Marion, 2009; Wimpey and Marion, 2010). On the biomass (FebruaryeApril) each year from 2014 to 2016.

Table 2 Physical attributes of blocks within Scottsdale's McDowell Sonoran Preserve.

Block Plant communitya Soil typeb Bedrockc Slope range Elevation range (m)

Gateway Ambrosia deltoidea e Parkinsonia microphylla Extremely Sandy Loam Metamorphic 20e29% (but 5% 605e772 mixed scrub association (154.121) along 100 m site control) Tom's Thumb Simmondsia chinensis e mixed scrub Very gravely sandy loam Granite 5e11% 774e901 association (154.123) Brown's Ranch Simmondsia chinensis e mixed scrub Very gravely sandy loam, Granite 0e3% 804e835 association (154.123) very gravely clay loam

a (Jones, 2015). b (USDA Natural Resource Conservation Service, 2017). c (Skotnicki, 2016). H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302 295

Fig. 1. Locations of transects in Scottsdale's McDowell Sonoran Preserve. 296 H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302

2.5. Data analyses Poisson Mixed Model (SAS Institute Inc., 2012). Tukey-Kramer was used for multiple comparisons tests. Volunteer count data was used to establish trail visitation levels. The 100 m site controls were analyzed with 2015 and 2016 data This data was log transformed to meet assumptions of a normal to test whether the 6 m distance from trails was a sufficient dis- distribution and analyzed using a mixed model with trail segment tance to control for trail effects. The response variables were tested as a fixed effect. Tukey's HSD was used to compare post hoc mul- using the same models described above. In each model plot posi- tiple comparisons. tion (trailside, 6 m, 100 m) and time as the covariate were included Richness and percent cover for native and nonnative plants and as fixed effects and block as a random effect. functional groups (native perennial and native annual plants) were All data analyses were generated using SAS/STAT software. © calculated. The three soil depth points in each trail segment end Copyright 2012 SAS Institute Inc. SAS and all other SAS Institute were averaged, thus there were two averaged data points per year Inc. product or service names are registered trademarks or trade- per transect (n ¼ 6 per transect). Differences between subsequent marks of SAS Institute Inc., Cary, NC, USA. Plots were generated in R years (2015e2014 and 2016e2015) were calculated using the (R Core Team, 2014) using GGPLOT2 (Wickham, 2009). averaged point for each transect end (resulting in two averaged differences for each transect and year pair, n ¼ 4 per transect). 3. Results Residual plots were used to evaluate the distributional properties of the data. We transformed the data to best meet assumptions of 3.1. Trail visitation normality and homogeneity of variance and to minimize outlier effects. Native species richness and cover, native perennial cover Trail visitation was significantly different (p < 0.0001, and richness, and native annual richness were square root trans- F(8,148) ¼ 59.38) and separated cleanly into three clear levels of use formed; native annual cover and soil depth differences were log based on Tukeys HSD (Table 3). Relative trail visitation levels within transformed. Data from three trailside e6 m plot pairs along a block were in agreement across the different estimation methods medium trail segment for 2015 and one 100 m site control for 2016 (mechanical and human counters; Table 3). were missing. Two plots with atypical rock cover were detected in the native plant richness data as outliers. We ran the analyses with 3.2. Soil crusts and without these plots and although the resulting significance of the tests did not change, the outliers were highly influential in the The loss of visible soil crusts on trailside plots compared with model and thus removed. These plots were not removed from other 6 m plots was mediated by the level of visitation along trails (sig- analyses as they had no effect. nificant interaction, Fig. 2, Table 4). Along low visitation trail seg- A repeated measures mixed model design with an autore- ments, there was no difference between soil crusts trailside and at gressive covariance error structure (AR (1)) with year, trail visita- 6 m, but along medium and high visitation trails, this difference tion level, plot position (trailside, 6 m), and the interaction between was significant (P < 0.0001, P ¼ 0.0002, respectively; Fig. 2). In the trail visitation and plot position as fixed effects, and block as a comparison with the 100 m site controls, both controls (6 m and random effect was used to evaluate native plant cover and richness, 100 m) supported significantly more soil crusts than the trailside including native annual and perennial functional groups. Soil depth plots (p < 0.0001). means were analyzed using the same model without (AR (1)) with year as a fixed effect and block as a random effect to test whether 3.3. Plant community response soil depth changed over time. Soil depth differences were also analyzed using the same model with (AR (1)), with year as Native plant cover and richness was reduced on trailside plots 2014e2015 and 2015e2016 and trail visitation levels as fixed ef- compared to the 6 m controls (Table 4). Responses of native plant fects, and block as a random effect to test whether soil depth cover and richness were most pronounced at the level of functional changed with trail visitation. Tukey's HSD was used to compare group (Tables 4 and 5), therefore in the figures we present plant post hoc multiple comparisons. We also ran the same analyses with cover and richness by functional group. Year as a covariate was the mean deviation from average rainfall for each year as a covar- significant for every plant response variable. iate instead of time. The model output was very similar between Native perennial richness and cover was higher on trailside plots the two approaches, indicating that year is largely a proxy for compared with 6 m plots, averaged over trail visitation level at the rainfall. We report analyses with time as a covariate because it (Table 4, Figs. 2 and 3). Native annual cover had a marginally sig- includes more factors associated with year other than total annual nificant interaction (Table 4), whereby annual plant cover was rainfall. higher in the trailside plots at low visitation levels, but in the higher Due to the lack of normality of nonnative cover, nonnative use plots cover was the same in trailside and 6 m plots (Fig. 2). richness and soil crust, alternative modeling techniques were Nonnative cover had a significant interaction in which trailside employed. These three variables had a high number of measured plots with low use had more nonnative cover than the 6 m plots, zero values and few unique percentages recorded. Thus, methods but the pattern switched at medium and high visitation levels typically used for count data, like Poisson regression, provided (Table 4, Fig. 2). Nonnative richness was highest on low use trails, more appropriate models. The values were rounded to the nearest and reduced on medium and high use trails averaged over distance integer for modeling. Similar to the standard mixed models, a from trail. (Table 4, Figs. 2 and 3). random effect for block that explicitly models the relationship In the comparisons including the 100 m site controls and only among the transects within a block was used. However, an autor- 2015 and 2016 data, the only plant parameters with significant plot egressive covariance error structure (AR (1)) term could not be position effects (trailside, 6 m, 100 m) were native plant richness, included. All models evaluated the impact of plot position and trail native perennial richness, and nonnative cover (Table 5). No dif- visitation level and their interaction controlling for year. Nonnative ferences in native plant richness were detected on plots 100 m from richness was modelled using a Poisson mixed model, the square trails compared with either the 6 m or trailside plots. But trailside root of soil crust was modelled using negative binomial mixed plots had lower perennial plant cover than 6 m (Tukey's adjusted model, and nonnative cover was modelled using a generalized P ¼ 0.03), consistent with the mixed model results without the site Poisson mixed model to account for overdispersion in the standard controls (Fig. 2). However native perennial richness on the 100 m H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302 297

Table 3 Mean visits per hour estimated by mechanical counter and volunteers for peak visitation. Capital letters denote difference between trail use based on volunteer counts using Tukey's HSD multiple means comparisons.

Block Trail Mechanical Count Volunteer Count Visitation level

N Mean S.E. N Mean S.E.

Tom's Thumb Tom's Thumb 9 59.61 7.23 19 64.37 6.22 HighA Gateway Gateway Saddle 7 46.07 9.00 15 61.13 6.80 HighA Gateway Windgate 6 15.83 1.35 17 23.82 2.86 MediumB Brown's Ranch Upper Ranch 2 26.00 0.50 18 23.06 2.60 MediumB Gateway Bell Pass 6 21.83 2.27 18 22.78 2.61 MediumB Tom's Thumb Marcus Landslide 9 19.89 4.99 19 19.42 3.14 MediumB Brown's Ranch Hackamore 3 14.50 3.40 18 13.69 1.43 MediumB Tom's Thumb Feldspar 1 3.00 19 3.32 0.79 LowC Brown's Ranch Rustler 9 0.61 0.40 17 2.44 0.39 LowC

plots was lower than the 6 m plots (Tukey's adjusted P < 0.01) and again consistent with the mixed model results without the site similar to the trailside plots. Trailside plots had reduced perennial controls (Fig. 2). As expected from the mixed model results without cover compared with the 6 m plots (Tukey's adjusted P ¼ 0.01), the site controls, trailside and 6 m nonnative plant cover did not

Fig. 2. Mean soil crust, native annual, native perennial, and nonnative percent cover by trail visitation level and plot position, averaged over year. Each error bar is constructed by using one standard error from the mean. Highest level significant differences are indicated by an asterisk for an interaction effect (*p < 0.05, **p < 0.01, ***p < 0.001). If no interaction, differences between treatments using Tukey's HSD multiple means comparisons are indicated by upper case letters for a trail visitation level effect (averaged over distance from trail) and lower case letters for differences between trailside and 6 m from trail plots (averaged over trail visitation). 298 H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302

Table 4 Full factorial mixed model analysis results for soil crust cover and percent plant cover and species richness of each functional group.

Plot position (trailside, 6 m) Trail visitation level (high, medium, low) Year (2014e2016) Position * Visitatio

df F P df F P df F P df F P

Soil crust cover 1757 31.60 <0.0001 2757 2.62 0.074 2757 0.99 0.374 2757 3.76 0.024 Native plant cover Richness 1791 6.12 0.014 2791 0.69 0.502 2791 72.97 <0.0001 2791 2.44 0.088 Cover 1793 4.81 0.029 2793 1.85 0.158 2793 9.84 <0.0001 2793 1.94 0.145 Native perennials Richness 1800 13.09 0.000 2800 0.05 0.956 2800 14.05 <0.0001 2800 1.13 0.322 Cover 1800 5.69 0.017 2800 0.11 0.892 2800 6.73 0.001 2800 1.78 0.169 Native annuals Richness 1800 0.12 0.733 2800 1.53 0.218 2800 25.52 <0.0001 2800 0.84 0.434 Cover 1763 0.76 0.384 2763 1.74 0.176 2763 13.10 <0.0001 2763 2.58 0.076 Nonnatives Richness 1793 0.11 0.735 2793 12.44 <0.0001 2793 0.62 0.538 2793 0.73 0.483 Cover 1793 0.00 0.992 2793 22.02 <0.0001 2793 21.25 <0.0001 2793 3.08 0.047

* sig p

Table 5 In areas next to trails perennial plant richness and cover were Mixed model analysis results for soil crust cover and percent plant cover and species reduced compared to the 6 m plots. Generally, trail construction of richness of each functional group for the comparisons with the 100 m site controls. established trails has a large initial effect (vegetation removal, soil Plot position (trailside, Year (2014e2016) disturbance) followed by lesser direct and indirect impacts (Cole, 6 m, 100 m) 1987; Leung and Marion, 1996). The reduction of perennial plants df F P df F P alongside trails in our study may be attributable to either initial trail building, subsequent trampling effects next to trails, or trail Soil crust cover 2617 23.71 <0.0001 1617 0.00 0.990 Native plants widening, but the effect was not driven by the different visitation Richness 2616 3.67 0.026 1616 52.3 <0.0001 levels studied. Nepal and Way (2007) used a similar experimental Cover 2617 0.82 0.440 1617 11.23 0.001 design comparing trailside and control plots by one high and one Native perennials low use trail in a Provincial Park, British Columbia, Canada. Similar Richness 2625 6.48 0.002 1625 1.45 0.229 to our results, they found that the high use trailside plots had less Cover 2625 1.45 0.235 1625 1.20 0.273 Native annuals vegetation and , , and fungi cover, and less species Richness 2625 0.4 0.668 1625 25.57 <0.0001 richness than control plots, and also that trailside plots on both Cover 2590 2.23 0.109 1590 15.32 0.000 high and low use trails had more exposed soil than the controls Nonnatives (Nepal and Way, 2007). In contrast to our results, others found that Richness 2617 0.73 0.480 1617 0.08 0.773 Cover 2617 5.15 0.006 1617 28.86 <0.0001 species richness increased in high use trailside plots (Bright, 1986; Hall and Kuss, 1989; Roovers et al., 2004a). This result is likely due * sig p

Fig. 3. Percent plant richness of functional groups by trail visitation and plot position. Each error bar is constructed by using one standard error from the mean. Highest level significant differences are indicated by an asterisk for an interaction effect (*p < 0.05, **p < 0.01, ***p < 0.001). If no interaction, differences between treatments using Tukey's HSD multiple means comparisons are indicated by upper case letters for a trail visitation level effect (averaged over distance from trail) and lower case letters for differences between trailside and 6 m from trail plots (averaged over trail visitation). an increase in annual plant richness or cover given the reduction in with alpine grassland, fen and bolster heath sites in Tasmania perennial cover in trailside plots, but annual plant cover was not (Whinam and Chilcott, 1999). Cole (1995a) results suggest that different between the trailside and 6 m control plots. In the instead of focusing on ecological systems, plant morphological Sonoran Desert, annuals’ limiting resource is moisture rather than characteristics may better explain response to trampling (Cole, light; winter annuals germinate in response to at least one 2.5 cm 1995a). Specifically, vegetation stature and erectness increases soaking rain event between October and January (Dimmitt, 2000), resistance, and matted grasses and sedges have the most tolerance and can benefit from moisture retention by perennial plants. In all to trampling, and broad leaved forbs the least (Cole, 1995a, 1995b; plant analyses, year, as a covariate which we found to be a good Monz et al., 2010). However, others found that upright woody proxy for rainfall, was a strong predictor of plant parameters. This shrubs and tall grasses were most vulnerable to trampling indicates that there is strong interannual variability explained (Whinam and Chilcott, 2003). In the arid southwest system studied largely by rainfall (Table 1). here, the annual plants which appear for only a short few months Several studies have tried to make generalizations about which after winter rains appear to be most resilient to trail effects. systems are most susceptible to trampling or trails, but the results range from densely forested vegetation in Oregon compared with 4.3. Nonnative plants open meadows or open forests (Cole, 1978); understory vegetation of mesophilous forests compared with heath and dry forests in We had expected that nonnative plants would be more abun- Belgium (Roovers et al., 2004b); floodplains and hemlock com- dant along the trail compared with the 6 m control plots and would munities compared with mesic and dry-mesic upland communities increase with visitation. Although an interaction response was in Indiana (Adkinson and Jackson, 1996); and shrublands compared detected in nonnative cover, the differences were very small 300 H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302 between trailside and the 6 m plots, and richness decreased for line (the route leading down any particular slope) (Marion and both trailside and 6 m plots nearby higher use trails, contrary to the Wimpey, 2017), suggesting that the design of trails in our study expected response. In the site control plots located 100 m from the may have been effective at minimizing erosion. nearest trail, nonnative cover was slightly higher than trailside and 6 m plots indicating that the nonnative species found in the trail 4.5. Site controls corridor do not appear to be distributed by trails. There were only six nonnative species found in our plots over the three years of the For the most part, the site control plots that were placed in each study, and two species were only found in one plot each and block 100 m away were similar to the 6 m plots, suggesting that 6 m another in nine plots. The three main nonnatives were red brome was a sufficient distance from the trail to serve as a control from (Bromus rubens), common Mediterranean grass (Schismus barba- disturbance associated with the trail. Native perennial richness tus), and redstem stork's bill (Erodium cicutarium), all of which are deviated from this pattern in which 100 m plots were more similar common throughout the MSP and other Phoenix area parks. Our to trailside than the 6 m plots. It is possible that because there was results, that nonnatives do not follow a pattern that indicate that only one 100 m site control per block, the site controls were too trails function as a primary conduit for nonnative plant spread into distant and different from the trailside plots to compare plant natural areas, are supported by studies comparing trailside and community characteristics. However, the pattern found with interior plots in the Rocky Mountains and foothills (Potito and nonnative plant cover using the site controls helps elucidate pat- Beatty, 2005; Tyser and Worleya, 1992). In contrast, however, a terns of nonnative plant dispersal as discussed above. similar study in Ontario, Canada showed a pattern consistent with spread from the trail outwards in which nonnative species were 5. Management implications highest along trails and decreased at three and 15 m from the trail (Patel, 2000). Other studies compare nonnative prevalence from Our results indicate that in this system trails concentrate high trailheads into the natural areas along trails finding a positive levels of use into a small area with little direct impact on sur- correlation with visitation use and nonnative plant species. These rounding protected natural areas as they are designed to do as long studies found that sites closer to the trailhead with higher use had as visitors remain “on-trail”. Careful design planning at the time of an increase of nonnative species cover (Potito and Beatty, 2005) and new trail construction can incorporate natural barriers that richness (Bella, 2011) compared with less travelled trail sections discourage off-trail use such as running trail segments alongside farther into the recreation area. Bella (2011) also found that the boulders or established growth whenever possible or using existing nonnative richness spread a longer distance from the trailhead on corridors like abandoned roads that already are incised into the high use trails. Although it's not clear from these studies to what terrain. To reduce the localized degradation of soil crusts on extent trails and trail visitation level contribute to the spread of established trails, managers may consider site management actions nonnative plants, in our system the two perennial plant species of such as physical barriers in conjunction with education campaigns, highest concern, buffelgrass (Pennisetum ciliare) and fountain grass which have been found to be effective in reducing off-trail use (Pennisetum setaceum), were not found in any of the study plots. (Hockett et al., 2017; Littlefair and Buckley, 2008; Park et al., 2008). This, together with the similarity between the trailside and 6 m In keeping with the natural aesthetic of the Preserve, we recom- control plots, suggest that trails are not a strong conduit for non- mend planting vegetation that is both consistent with the sur- natives in the McDowell Sonoran Preserve at the time of the study. rounding plant community and also of sufficient size or with characteristics that will discourage off-trail use. For example, in the 4.4. Soil erosion Sonoran Desert buckhorn cholla (Cylindropuntia acanthocarpa)or teddy bear cholla (Cylindropuntia bigelovii) plants or segments are Surprisingly, there was no detectable change in soil erosion in common local species that can be easily transplanted and create a the trails as determined by repeated trail-depth measurements, prickly barrier for visitors. regardless of trail visitation level. This was unexpected since the Trail maintenance can also reduce or eliminate problems within trails included in the study had markedly different geological and a trail that may be causing visitors to go off-trail to avoid obstacles. tread characteristics, from erosion-prone granitic grus (sand and Signage to stay on the trail can be effective in combination with small gravel) in the Brown's Ranch and Tom's Thumb blocks to revegetation (Hockett et al., 2017). New interpretive displays at erosion-resistant exposed metamorphic rock and compacted fine trailheads can explain the effects of trail widening on soil crusts and dirt in the Gateway block (Skotnicki, 2016), as well as different encourage visitors to stay within the trail boundary and “don't bust usage levels within and between blocks. Our results are contrasted the crust”. Because these trails are multi use, recommendations for by another study conducted in the arid southwest (Guadalupe horse riders could include choosing wider trails and for mountain Mountains National Park, Texas, USA) in which most measured sites bikers to be especially vigilant about keeping their turning radius showed active soil movement over a 17-month period in an area within the established trail. Etiquette tips for allowing other users composed primarily of limestone, especially on major trails to pass could include trying to stay on the trail if possible or care- compared with control transects (Fish and Brothers, 1981). These fully choosing a rock or soil crust free area off trail to stand. Since changes appeared to be due primarily to sediment deposition personal contacts may be more effective than interpretive signs in rather than erosion, as the trails intercepted waterborne sediment changing off trail use (Hockett et al., 2017; Littlefair and Buckley, (Fish and Brothers, 1981). Several studies suggest that trail incision 2008), managers and volunteers could be trained and encouraged depth and rates are highly influenced by trail design (trail grade to promote the message. and trail-slope angle) as shown in South Carolina, USA (Beeco et al., 2013), western Australia (Gager and Conacher, 2001), and western 6. Conclusions Tasmania (Dixon et al., 2004), and only moderately influenced by trail use (Beeco et al., 2013; Dixon et al., 2004). Beeco et al. (2013) Monz et al. (2013) reviewed the relationship between recreation found that horse traffic had similar influence over trail erosion as use and impact on vegetation and soil. The predominant model is trail design. In a study of erosional loss in relation to trail design, the asymptotic and curvilinear, whereby in an undisturbed area with highest levels of erosion were associated with either very steep low use levels, small differences in the amount of visitation can lead trails or steep trails that also were aligned close to the natural fall to substantial impact, whereas in already disturbed areas with high H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302 301 use (trails or campsites), additional impact has proportionally less 335e362. effect (Monz et al., 2013). An alternative model suggests a sigmoidal Boucher, D.H., Aviles, J., Chepote, R., Domínguez Gil, O.E., Vilchez, B., 1991. Recovery of trailside vegetation from trampling in a tropical rain forest. Environ. response in which at low levels of use, there is little impact, but at Mutagen. 15, 257e262. https://doi.org/10.1007/BF02393857. medium use, there is a steep impact curve that flattens again after a Bright, J.A., 1986. Hiker impact on herbaceous vegetation along trails in an ever- e secondary threshold has been reached (Monz et al., 2013). In our green woodland of Central Texas. Biol. Conserv. 36, 53 69. https://doi.org/ 10.1016/0006-3207(86)90101-1. system, only soil crust followed the sigmoidal response; plant cover Brooks, M.L., Lair, B., 2005. Ecological Effects of Vehicular Routes in a Desert and richness in trailside plots did not show significant trends ac- Ecosystem. USA US Dep, Las Vegas, NV. Inter. … 23. cording to use level indicating that either additional impact beyond Brown, D.E., Lowe, C.H., Pase, C.P., 1979. A digitized classification system for the biotic communities of North America, with community (series) and associated initial trail building has little effect, or that these trails have not examples for the Southwest. J. Arizona- Acad. Sci. experienced high enough use or disturbance to reach a first Cole, D., 1978. Estimating the susceptibility of wildland vegetation to trailside threshold. The shape of the curve has important management alteration. J. Appl. Ecol. 15, 281e286. Cole, D., 1987. Research on soil and vegetation in wilderness: a state-of-knowledge implications. If trail visitation continues to grow in the MSP, it review. General Technical Report INT-220. In: Lucas, R.C. (Ed.), Proceedings- would be useful to know whether to expect a second threshold or National Wilderness Research Conference: Issues, State-of- Knowledge, Future continued minimal response. As discussed earlier, trail impacts are Directions. Forest Service, Intermountain Research Station, Fort Collins, CO, e expected to differ by ecological system because environmental pp. 135 177. Cole, D.N., 1995a. Experimental trampling of vegetation. II. Predictors of resistance factors such as climate and geology and the intermediate elements and resilience. J. Appl. Anim. Res. 32, 215e224. of topography, soil, and vegetation type significantly affect the Cole, D.N., 1995b. Experimental trampling of vegetation. I. Relationship beteween e degree and type of trail degradation (Leung and Marion, 1996). For trampling intensity and vegetation response. J. Appl. Ecol. 32, 203 204. Cole, D.N., 1990. Trampling disturbance and recovery of cryptogamic soil crusts in the first time, we were able test how plants and soil crusts respond Grand Canyon National Park. Gt. Basin Nat. Nat. 50, 321e325. to trail proximity and different trail visitation levels, as well as test Crisfield, V.E., Macdonald, S.E., Gould, A.J., 2012. Effects of recreational trafficon soil erosion on trails in a Sonoran Desert setting. We found that soil alpine plant communities in the northern Canadian Rockies. Arctic Antarct. Alpine Res. 44, 277e287. https://doi.org/10.1657/1938-4246-44.3.277. crusts are an early indicator of disturbance, and that perennial and D'Antonio, A., Monz, C., Larson, N., Rohman, A., 2016. An application of recreation annual plants and soil erosion appear to be resilient to change at resource assessment techniques to inform management action in an urban- least at current levels of use. Trails did not appear to be an proximate natural area. J. Outdoor Recreat. Tour. 14 https://doi.org/10.1016/ j.jort.2016.04.002. important conduit for nonnative species dispersal into the MSP. Dimmitt, M.A., 2000. Plant ecology of the sonoran desert region. In: Phillips, S.J.W.C.P. (Ed.), A Natural History of the Sonoran Desert. Arizona- e Acknowledgements Desert Museum Press, Tucson, AZ, pp. 129 152. Dixon, G., Hawes, M., McPherson, G., 2004. Monitoring and modelling walking track impacts in the Tasmanian wilderness world heritage area, Australia. J. Environ. We thank the McDowell Sonoran Conservancy and an anony- Manag. 71, 305e320. https://doi.org/10.1016/j.jenvman.2004.02.006. mous donor for financial support of this project. A special thanks Fish, B., Brothers, L., 1981. NATIONAL which is largely responsible for the uniqueness of this site. The driest from November through April ( 70 % of the annual goes to L. Miller for developing the citizen science training and precipitation occurs from May through October). Annual precipitation ranges structure for this project. Citizen scientists K. Fieldstad, B. Pringle, J. from 340 mm 8. Millavec, J. McEnroe, B. Sowle, C. Holt and L. Rivera provided Flood Control District of Maricopa County, 2017. Climate Data [WWW Document]. http://alert.fcd.maricopa.gov/alert/Google/v2/raininfo.html. (Accessed 5 essential project leadership and K. Barron and S. Doan shared their February 2017). botanical expertise. Thank you to the additional 60 citizen scientists Gager, P., Conacher, A., 2001. Erosion of access tracks in Kalamunda National Park, who helped with field data collection and data entry. The City of Western Australia: causes and management implications. Aust. Geogr. 32, 343e357. https://doi.org/10.1080/00049180120100068. Scottsdale provided permission for conducting this study in the Hall, C.N., Kuss, F.R., 1989. Vegetation alteration along trails in Shenandoah National McDowell Sonoran Preserve. Park, Virginia. Biol. Conserv. 48, 211e227. https://doi.org/10.1016/0006- 3207(89)90119-5. Hammitt, W.E., Cole, D.N., 1998. Wildland Recreation: Ecology and Management, References second ed. John Wiley and Sons, New York. Harper, K.T., Marble, J.R., 1988. A role for nonvascular plants in management of arid Adkinson, G.P., Jackson, M.T., 1996. Changes in ground-layer vegetation near trails in and semiarid rangelands. In: Tueller, P.T. (Ed.), Vegetation Science Applications midwestern US forests. Nat. Area J. 16, 10e23. for Rangeland Analysis and Management. Kluwer Academic Publisher, Dor- Allen, C.D., 2009. Monitoring environmental impact in the upper sonoran lifestyle: drecht, Germany, pp. 135e169. a new tool for rapid ecological assessment. Environ. Mutagen. 43, 346e356. Hockett, K.S., Marion, J.L., Leung, Y.F., 2017. The efficacy of combined educational https://doi.org/10.1007/s00267-008-9212-5. and site management actions in reducing off-trail hiking in an urban-proximate Andres-Abell an, M., Lopez-Serrano, F.R., Morote, F.A.G., Del Cerro-Barja, A., 2006. protected area. J. Environ. Manag. 203, 17e28. https://doi.org/10.1016/ Assessment of trampling simulation impacts on native vegetation in Mediter- j.jenvman.2017.06.073. ranean sclerophyllous forest. Environ. Monit. Assess. 120, 93e107. https:// Johansen, J.R., 1993. Cryptogamic crusts of semiarid and arid lands of North doi.org/10.1007/s10661-005-9051-2. America. J. Phycol. 29, 140e147. Ballantyne, M., Pickering, C.M., 2015. The impacts of trail infrastructure on vege- Jones, S., 2015. Plant associations of Scottsdale's McDowell sonoran preserve: a tation and soils: current literature and future directions. J. Environ. Manag. report to the city of Scottsdale doi:MSCFI-MSP-008-R. https://doi.org/10.1016/j.jenvman.2015.08.032. Jones, S., Hull, C., 2014. Vegetation and flora of the McDowell sonoran preserve. Beeco, J.A., Hallo, J.C., English, W.R., Giumetti, G.W., 2013. The importance of spatial Canoma 10, 1e34. nested data in understanding the relationship between visitor use and land- Kuss, F.R., Graefe, A.R., Vaske, J.J., 1990. Visitor impact management: a review of scape impacts. Appl. Geogr. 45, 147e157. https://doi.org/10.1016/ research. Natl. Park. Conserv. Assoc. 256. j.apgeog.2013.09.001. Leung, Y., Marion, J.L., 2000. Recreation impacts and management in Wilderness: a Bella, E.M., 2011. Invasion prediction on Alaska trails: distribution, habitat, and trail state-of-knowledge review the field of recreation ecology. In: Cole, David N., use. Invasive Plant Sci. Manag. 4, 296e305. https://doi.org/10.1614/IPSM-D-10- McCool, Stephen F., Borrie, W.T.O., Jennifer, C. (Eds.), Wilderness Science in a 00083.1. Time of Change Conferenced Volume 5: Wilderness Ecosystems, Threats, and Belnap, J., 1998. Choosing indicators of natural resource condition: a case study in Management. U.S. Department of Agriculture, Forest Service, Rocky Mountain , Utah, USA. Environ. Mutagen. 22, 635e642. https:// Research Station, pp. 23e48. Yu-Fai Leung is Assistant Professor, Department of doi.org/10.1007/s002679900135. Parks, Recreation and, Missoula, MT. Belnap, J., 1994. Potential value of cyanobacterial inoculation in revegetation efforts. Leung, Y.F., Marion, J.L., 1996. Trail degradation as influenced by environmental In: Monsen, S.B., Kitchen, S.G., P.-E., M., A.R (Eds.), Ecology and Management of factors: a state-of-the-knowledge review. J. Soil Water Conserv. 130e136. Annual Rangelands. U.S. Department of Agriculture, Forest Service, Ogden, Littlefair, C., Buckley, R., 2008. Interpretation reduces ecological impacts of visitors Utah, pp. 179e185. Technical Report INT-GTR-313. to world heritage site. AMBIO A J. Hum. Environ. 37, 338e341. https://doi.org/ Belnap, J., Gardner, J., 1993. Soil microstructure in soils of the Colorado Plateau: the 10.1579/07-R-393.1. role of the cyanobacterium Microcoleus vaginatus. West. North Am. Nat. 53, Marion, J.L., Leung, Y.-F., 2001. Trail resource impacts and an examination of alter- 40e47. native assessment techniques. J. Park Recreat. Adm. 19, 17e37. Belnap, J., Gillette, D.A., 1997. Disturbance of biological soil crusts: impacts on po- Marion, J.L., Wimpey, J., 2017. Assessing the influence of sustainable trail design and tential wind erodibility of sandy desert soils in SE. Land Degrad. Dev. 8, maintenance on soil loss. J. Environ. Manag. 189, 46e57. https://doi.org/10.1016/ 302 H.I. Rowe et al. / Journal of Environmental Management 207 (2018) 292e302

j.jenvman.2016.11.074. and vegetation recovery in some forest and heathland communities. Appl. Veg. McDowell Sonoran Conservancy, 2014. The Flora and Fauna of Scottsdale's McDo- Sci. 7, 111e118. https://doi.org/10.1658/1402-2001(2004)007{[}0111:ETAVRI] well Sonoran Preserve. 2.0.CO;2. ® ® Metting, B., 1991. Biological Surface Features of Semiarid Lands and Deserts. Marcel SAS Institute Inc, 2012. SAS/STAT 9.3 User's Guide [WWW Document]. SAS/STAT Dekker, Inc, New York. 9.3 User's Guid. https://support.sas.com/documentation/cdl/en/statug/63962/ Monz, C.A., Cole, D.N., Leung, Y.F., Marion, J.L., 2010. Sustaining visitor use in pro- HTML/default/viewer.htm#statug_glimmix_sect028.htm. (Accessed 6 June tected areas: future opportunities in recreation ecology research based on the 2017). USA experience. Environ. Mutagen. 45, 551e562. https://doi.org/10.1007/ Schimel, D.S., Kelly, E.F., Yonker, C., Aguilar, R., Heil, R.D., 1985. Effects of erosional s00267-009-9406-5. processes on nutrient cycling in semiarid landscapes. In: Caldwell, D.E., Monz, C.A., Pickering, C.M., Hadwen, W.L., 2013. Recent advances in recreation Brierley, J.A., Brierley, C.L. (Eds.), Planetary Ecology. Van Nostrand Reinhold, ecology and the implications of different relationships between recreation use New York. and ecological impacts. Front. Ecol. Environ. 11, 441e446. https://doi.org/ Skotnicki, S.J., 2016. Geology of the McDowell Mountains. Arizona Geological Survey 10.1890/120358. Contributed Map CM-16-A, 1 map plate. http://tinyurl.com/GeoMap-Mcdowe Mueller-Dombois, D., Ellenberg, H., 1974. Aims and Methods of Vegetation Ecology. llMtns. John Willey, New York. Tyser, R.W., Worley, C.A., 1992. Alien flora in grasslands to road and trail adjacent in Nepal, S.K., Way, P., 2007. Comparison of vegetation conditions along two back- glacier National corridors (U. S. A.). Conserv. Biol. 6, 253e262. https://doi.org/ country trails in mount Robson Provincial Park, British Columbia (Canada). 10.1007/BF00410220. J. Environ. Manag. 82, 240e249. https://doi.org/10.1016/j.jenvman.2005.12.016. USDA Natural Resource Conservation Service, 2017. Web Soil Survey [WWW Olive, N.D., Marion, J.L., 2009. The influence of use-related, environmental, and Document]. https://websoilsurvey.sc.egov.usda.gov/App/HomePage.htm. managerial factors on soil loss from recreational trails. J. Environ. Manag. 90, (Accessed 1 January 2014). 1483e1493. https://doi.org/10.1016/j.jenvman.2008.10.004. Webb, R.H., 1983. Compaction of desert soils by off-Road ve. Environ. Eff. off-road Park, L.O., Manning, R.E., Marion, J.L., Lawson, S.R., Jacobi, C., 2008. Managing visitor Veh. Impacts Manag. Arid Reg. 51e79. impacts in Parks: a multi-method study of the effectiveness of alternative Whinam, J., Chilcott, N., 1999. Impacts of trampling on alpine environments in management Practices. J. Park Recreat. Adm. 26, 97e121. central Tasmania. J. Environ. Manag. 57, 205e220. https://doi.org/10.1006/ Patel, A., 2000. Assessing the impacts of deer browsing, prescribed burns, visitor jema.1999.0302. use, and trails on an oak-pine forest: Pinery Provincial Park, Ontario. Canada. Whinam, J., Chilcott, N.M., 2003. Impacts after four years of experimental trampling Nat. Areas J. 20, 250e260. on alpine/sub-alpine environments in western Tasmania. J. Environ. Manag. 67, Potito, A.P., Beatty, S.W., 2005. Impacts of recreation trails on exotic and ruderal 339e351. https://doi.org/10.1016/S0301-4797(02)00218-9. species distribution in grassland areas along the Colorado Front Range. Environ. White, E.M., Bowker, J.M., Askew, A.E., Langner, L.L., Arnold, J.R., English, B.K., 2014. Mutagen. 36, 230e236. https://doi.org/10.1007/s00267-003-0177-0. Federal Outdoor Recreation Trends: Effects on Economic Opportunities October R Core Team, 2014. R: a Language and Environment for Statistical Computing. 2014. Rickard, C.A., McLachlan, A., Kerley, G.I.H., 1994. The effects of vehicular and Wickham, H., 2009. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag, pedestrian traffic on dune vegetation in South Africa. Ocean Coast Manag. 23, New York. 225e247. https://doi.org/10.1016/0964-5691(94)90021-3. Williams, J.D., Dobrowolski, J.P., West, N.E., Gillette, D.A., 1995. Microphytic crust Roovers, P., Baeten, S., Hermy, M., 2004a. Plant species variation across path eco- influences on wind erosion. Trans. Am. Soc. Agric. Eng. 38, 131e137. tones in a variety of common vegetation types. Plant Ecol. 170, 107e119. https:// Wimpey, J.F., Marion, J.L., 2010. The influence of use, environmental and managerial doi.org/10.1023/B: VEGE.0000019028.89623.26. factors on the width of recreational trails. J. Environ. Manag. 91, 2028e2037. Roovers, P., Verheyen, K., Hermy, M., Gulinck, H., 2004b. Experimental trampling https://doi.org/10.1016/j.jenvman.2010.05.017.